Equity duration in China: A deep learning approach

We propose a method to estimate equity duration using convolutional neural network. We find that the duration long–short portfolio annual returns based on this duration that we estimate in Chinese stock market are 20.43% for equal-weighted and 18.24% for value-weighted respectively. We empirically demonstrate that this long–short portfolio return is: (i) strong even after firm size is controlled; and (ii)significant after risk is adjusted. Then, we form a new value factor based on the duration, which subsumes the book-to-market ratio and price-earnings ratio. We shed light on the equity duration in Chinese market and the application of neural network in estimating firm cash flow and equity duration.